Identity Verification in Smartphones as Social Intersectionality: Inclusive Design of Contactless Fingerprints to Mitigate Skin Tone and Gender Bias

The researcher will develop a contactless biometric mobile security application to correct the vulnerabilities of deep AI and optical sensors and allow marginalized people equal access to data security regardless of gender and  skin tone.

Funded by the CCI Northern Node

Project Investigator

Principal Investigator (PI): Emanuela Marasco, George Mason University's Center for Secure Systems.

Rationale and Background

Marginalized people face technological discrimination due to the inability of AI security systems using optical sensors to identify salient features. Current optical sensors, including smartphone cameras, are limited in their ability to interpret  variations in skin tone and gender-related features. For example, users with darker skin tones tend to experience a lower level of accessibility to these security systems.  

Methodology

The researcher will identify the impact  of physical vulnerabilities, then use the findings to retrain AI models to mitigate issues and protect all users by:

• Modeling the optical properties of finger skin and categorizing finger photo images of individuals.
• Conducting a fairness evaluation of finger photo technology by analyzing the impact of skin tone and gender on matching and presentation attack detection (PAD) algorithms.
• Enhancing inclusiveness by retraining recent deep learning architectures to work well with RGB (red, green, blue) images and other color spaces.
• Interpreting improved inclusivity via Explainable AI. 

Projected Outcomes

The researcher will create an equitable contactless fingerprint verification technology (utilizing finger photos and videos), leading to higher standards of biometric security and equal accessibility. 

Application of this technology will allow greater and low-cost access to highly  secure biometric identification for people with marginalized identities.